Cancer screening cadence

A handful of cancer screening tests genuinely save lives in the people they were designed for. Several others mostly find cancers that would never have caused harm — so the useful question is not how much screening you do, but which tests, and when.

For two generations the framing was simple: more screening, earlier, for everyone, saves lives. The 2022–2026 evidence has not overturned that picture so much as compressed it. A few modalities — biennial mammography from 40, colonoscopy or stool testing from 45, primary human papillomavirus (HPV) testing from 30, low-dose CT in heavy ever-smokers, one-time hepatitis C and abdominal aortic aneurysm screens in defined groups — sit on robust evidence of preventing cancer or cancer death. The rest carry net benefit only in narrow risk strata, and a handful of routine tests now have a documented overdiagnosis problem that rivals or exceeds their mortality benefit. The largest randomized cancer-screening meta-analysis of the modern era found that across six common modalities, mean life-extension is measured in days, not years. That is the calibration this article is built around.

The first reframe: disease-specific vs all-cause mortality

A screening test can prevent death from cancer X without measurably extending the average life of the screened population. The 2023 JAMA Internal Medicine meta-analysis of 18 randomized trials in 2.1 million participants quantified exactly how compressed that gain is when averaged across everyone screened, including those who never had the disease and those who avoid cancer death only to die of something else.[1]

ModalityMean lifetime gained per person screenedRange compatible with the data
Sigmoidoscopy110 days0 to 274 days
Low-dose CT (lung)107 days286 days lost to 430 gained
Prostate-specific antigen (PSA)37 days37 lost to 73 gained
Colonoscopy37 days146 lost to 146 gained
Mammography (biennial 50–69)0 days190 lost to 237 gained
Fecal occult blood test0 days71 lost to 71 gained

The third column is the 95% confidence interval — the span in which the true effect most plausibly lies. Every row but the first includes the possibility of losing time as well as gaining it, and the first stops exactly at zero. That is the real finding: not that these tests buy a few weeks, but that the trials are compatible with their buying nothing.

These are means, and a mean of zero is compatible with substantial individual benefit — the woman whose Stage I tumor was caught at mammography gains years; the millions of women who never develop breast cancer gain none. But the population frame is the one the policy decision sits on, and it argues for a small number of high-value tests done well, not a long list done because they exist.

A reasonable counter-fear — that screening kills patients indirectly, through false-positive cascades, radiation, or unnecessary procedures — was tested in a 2025 JAMA Network Open meta-analysis of 17 randomized trials in 1.3 million participants. Deaths from non-cancer causes in the screened arms rose by 0.2% (95% CI −0.5% to 0.9%), a difference indistinguishable from zero — and the interval rules out any increase larger than about 1%.[2] Screening doesn't passively harm. It just earns less than its marketing implies.

Colorectal — start at 45, complete what you start (Strong)

The US Preventive Services Task Force (USPSTF) lowered the colorectal cancer (CRC) screening initiation age from 50 to 45 in May 2021, citing a near-15% rise in colorectal adenocarcinoma among 40–49-year-olds between 2000–2002 and 2014–2016. Grade B from 45 to 49; Grade A from 50 to 75; individualized 76–85.[3] Incidence in the under-50s is still climbing — 3% a year in adults aged 20 to 49 across 2013–2022.[4] Clinicians adopted it fast: in one large US health system, screening orders for 45–49-year-olds had nearly caught up with those for 50-year-olds within five months of the change.[5]

The interesting evidence is NordICC, the first large randomized controlled trial (RCT) of screening colonoscopy itself, which randomly allocated healthy 55–64-year-olds in Poland, Norway, Sweden, and the Netherlands, one to screening for every two left alone; follow-up data were analysed for 84,585 of them, in Poland, Norway, and Sweden.[6] After 10 years, the intention-to-screen analysis showed CRC incidence of 0.98% in the invitation arm vs 1.20% in usual care — an 18% relative reduction. CRC-specific mortality was 0.28% vs 0.31%, not statistically significant.

Thirteen-year results followed in 2026, on the 84,583 participants from Norway, Poland, and Sweden whose follow-up could be analysed. The incidence benefit held and sharpened: 1.46% versus 1.80%, a 19% relative reduction, rising to 45% in a per-protocol model that assumes everyone invited had been screened. Cancer deaths moved much less — 0.41% versus 0.47%, a difference compatible with no effect, and still not significant under that same perfect-adherence model.[7] Two details are worth carrying into a decision. The benefit sits almost entirely in the descending colon and rectum, with essentially no effect on cancers higher up, and it was clear in men but not in women. And the trial was underpowered for its mortality question by construction: colorectal cancer killed 0.47% of the unscreened arm against the 0.82% its designers had assumed, so the events needed to detect a realistic mortality effect were never there.

The signal-to-noise problem was adherence: only 42% of those invited actually showed up for the colonoscopy. The honest reading is that colonoscopy prevents colorectal cancer when it happens, that its effect on dying of colorectal cancer remains unproven after thirteen years, and that the dropout rate and the missing deaths both work against the trial. On lifespan there is no signal at all: death from any cause was 11.03% in the invited group and 11.04% in usual care — the disease-specific gain gets absorbed by competing mortality in the 55–75 window.[8]

What this means in practice. Start at 45. Use the modality you will actually complete — that is not a consolation prize, it is the finding. COLONPREV randomised 57,404 Spanish adults aged 50 to 69 to an invitation to a one-time colonoscopy or to a stool test every two years, and after 10 years colorectal cancer mortality was 0.22% against 0.24%, a difference of two hundredths of a percentage point that met its pre-specified test for non-inferiority.[9] The mechanism is the article's own thesis: more people did the stool test than showed up for the colonoscopy, 39.9% against 31.8%. And it is not simply that turnout rescued the weaker test: in a post-hoc analysis of the 17,270 people who actually started screening, 10-year colorectal cancer incidence and death from any cause were statistically indistinguishable either way.[10] Adherence is the variable that actually moves the outcome.

Breast — biennial from 40, with one open question (Strong)

The USPSTF moved the initiation age for screening mammography from 50 to 40 in April 2024, Grade B for biennial screening through age 74.[11] The driver was a 2%/year rise in invasive breast cancer incidence in the 40–49 cohort between 2015 and 2019, plus modeling that starting at 40 rather than 50 averts an additional ~1.3 breast cancer deaths per 1,000 women screened over a lifetime — and an additional ~1.8 per 1,000 Black women.[12] The larger increment sits on a larger total: screening biennially from 40 to 74 averts 10.7 deaths per 1,000 Black women against 8.2 per 1,000 women overall, and Black women's breast cancer mortality is about 40% higher than White women's.[13]

Two clinical scenarios still carry an "I" statement — insufficient evidence rather than affirmative recommendation:

  • Screening beyond age 75. No trial data; competing mortality starts to dominate.
  • Supplemental ultrasound or MRI for dense breasts (more than 40% of screened women). Extra imaging undeniably finds more tumors, but not equally: in a randomized comparison in women with dense breasts and a normal mammogram, abbreviated MRI and contrast-enhanced mammography each found roughly 17 to 19 additional cancers per 1,000 examinations, against about 4 for automated whole-breast ultrasound — though contrast-enhanced mammography was the only arm with a real harm signal, occasional reactions to the iodine contrast dye.[14] No trial has yet shown that any of this translates into fewer breast cancer deaths.

The DCIS overdiagnosis problem

Ductal carcinoma in situ (DCIS) — "Stage 0" — accounts for about 18% of newly diagnosed breast cancers in the US, almost all of it detected only because of screening.[15] A micro-simulation model of biennial screening from age 50 estimates that roughly 1 in 5 detected DCIS cases is overdiagnosis — 20%, with a plausible range of 13–30%, and little variation by tumor grade.[16] The historical response to any DCIS has nonetheless been immediate surgery.

What happens when that surgery is not done is now measurable. An observational cohort of 1,780 US women whose DCIS was not operated on within six months found invasive cancer in the same breast in 8–14% of them at eight years — 8.5% in the low-risk group, 13.9% in the high-risk group.[17] That is far below what the older pathology series implied: in the historical cohorts of untreated DCIS that justified routine lumpectomy plus radiation plus endocrine therapy, 36–39% of women developed invasive cancer in the same breast, though over a median follow-up of more than 30 years.[18] Eight-year survival free of breast-cancer death was 96.4% overall and 98.1% in the low-risk group.[19] The COMET randomized trial has now tested directly whether operating on low-risk DCIS is necessary at all. It enrolled 995 women with hormone-receptor-positive, grade 1 or 2 ductal carcinoma in situ and analysed 957, randomly allocated to monitoring every six months or to guideline-concordant surgery with or without radiation. After two years, invasive cancer had appeared in the same breast in 4.2% of the monitored group and 5.9% of the operated group — a difference of −1.7 percentage points whose upper bound, 0.95%, sits well inside the 5-point margin the trial set for calling monitoring no worse. Quality of life was much the same either way.[20], [21] Two years is short for a disease measured in decades, and the trial cannot say what happens at ten. But the direction of the result is that low-risk DCIS did not punish the women who waited.

The site's overall framing applies: screening doesn't cause cancer, but it does manufacture cancer diagnoses, and that distinction matters for what you do with a "Stage 0" result.

Cervical — primary HPV, every five years (Strong)

Primary testing for high-risk strains of human papillomavirus (HPV) detects 40% more high-grade precancerous lesions — grade 3 cervical intraepithelial neoplasia or worse, the stage just short of invasion — and 30% more cervical cancers per screening round than liquid-based cytology.[22] In the round after a negative HPV test, the odds of such a lesion are about four times lower than after negative cytology — an earlier, much smaller analysis of the same programme put it near sevenfold.[23], [24] On that basis the UK, the Netherlands, and the European Commission's screening guidance extend the interval to 5 years for HPV-negative women (longer in some models for women 40+).[25]

The behavioral catch is that women habituated to a 3-year cytology cadence often experience the longer interval as undertreatment. The biology argues the other way — persistent HPV infection takes years to progress through cellular dysplasia to invasive cancer, and a negative HPV test has a very high forward negative predictive value. Patient education is the actual implementation problem; the science isn't.

Two exceptions change the cadence, and both are easy to miss. A self-collected sample buys a shorter interval, not the same one. Self-collection was approved in the US in 2024 for use in a clinical setting and in 2025 for an at-home device, and the American Cancer Society now treats a self-collected vaginal specimen as acceptable for average-risk people aged 25 to 65 — but recommends repeating it after 3 years, against 5 for a clinician-collected test.[26] And clearing an HPV infection does not reset you to the five-year clock. Women who test positive and then negative carry about three times the subsequent risk of a high-grade lesion compared with women who were negative throughout, which is why the study behind the five-year interval concludes that this group should stay on three years.[27]

Prostate — risk-stratified PSA, not population PSA (Moderate)

The two-decade transatlantic argument between ERSPC (mortality benefit) and PLCO (no benefit) has largely been resolved by recognizing that PLCO's control arm was massively contaminated by opportunistic prostate-specific antigen (PSA) testing outside the trial.[28] Adjusting for mean lead time — the average amount of time by which screening brings a diagnosis forward — makes the two trials roughly compatible, with PSA screening conferring a 7–9% reduction in prostate cancer death per year of lead time.

At 23 years of follow-up across 162,236 men, ERSPC shows 13% lower prostate cancer mortality in the screened group (rate ratio 0.87, 95% CI 0.80–0.95).[29] This interval stays below 1.0; for a ratio like this one, an interval that crossed 1.0 would be compatible with no effect at all. In absolute terms the benefit is much smaller than "13%" sounds: about 2 fewer prostate cancer deaths per 1,000 men invited over 23 years. To prevent one prostate cancer death, about 456 men must be invited to screen and 12 must be diagnosed and treated — both figures uncertain enough to matter, with plausible ranges of 306 to 943 and 8 to 26. The overdiagnosis the next paragraph argues about has a number too: men in the screened group were diagnosed with prostate cancer 30% more often than men who were not screened, while their risk of dying of it fell by only about 2 per 1,000. The trade does improve with time: at 16 years it took 628 invitations and 18 diagnoses to prevent one death, against 456 and 12 at 23 years.

The implication is not "PSA doesn't work." It is that untargeted PSA testing on every adult man generates a large overdiagnosis and overtreatment cost — incontinence, erectile dysfunction, the cardiovascular morbidity of androgen deprivation — for a small absolute mortality gain. Modern urological guidelines and several European national programs are converging on a stratified pathway: baseline PSA, longer intervals for low values, MRI before biopsy, and active surveillance for Gleason 6 disease. That pathway has now been tested in randomised trials — ProScreen, which put PSA, a blood-marker panel and MRI ahead of any biopsy, found, in a preliminary report of its first screening round, about 4.6 extra aggressive cancers for every extra low-grade one[30] — and in March 2026 the UK National Screening Committee recommended the UK's first prostate programme, but only for men 45–61 with a pathogenic BRCA2 variant and a family history, explicitly declining population screening. Discuss the test; don't reflexively order it.

Lung — low-dose CT in actual heavy smokers (Strong)

The USPSTF 2021 update broadened low-dose CT (LDCT) eligibility to annual screening for adults 50–80 with a history of at least 20 pack-years — one pack a day for 20 years, or the equivalent — who currently smoke or quit within 15 years.[31] Screening stops once a person has been off cigarettes for 15 years, or develops a health problem that substantially limits life expectancy or the ability or willingness to undergo curative lung surgery.

That 15-year cutoff is contested. The American Cancer Society's 2023 guideline keeps the age band and the 20 pack-year threshold but removes years-since-quitting entirely — it is "not an eligibility criterion to begin or to stop screening".[32] The reason is that absolute lung cancer risk keeps climbing past 15 quit-years, by roughly 9% a year, because ageing outruns the benefit of having stopped.[33] US insurance follows the Task Force, not the Society, so someone who quit 18 years ago after 40 pack-years is eligible on one reading and unfunded on the other.

LDCT works because it catches small, operable nodules years before symptoms. In the National Lung Screening Trial, three annual scans cut lung cancer deaths by 20% — but against annual chest X-ray, not against no screening at all.[34] The Dutch–Belgian NELSON trial, which did compare against no screening, found about 24% fewer lung cancer deaths in men at ten years; its 2,594-woman subgroup was too small to settle the question in women.[35]

False positives are common (benign granulomas and hamartomas light up the same way), and serial surveillance scans, bronchoscopy, and needle biopsy of benign tissue are part of the cost — which is why the eligibility criteria are tight. Screening buys a year of life at a defensible price in the target population. The economic analysis inside the trial put it at about $81,000 per quality-adjusted life-year, with a plausible range running from $52,000 to $186,000 — the top of which is above what US payers usually accept. Its authors were explicit that "modest changes in our assumptions would greatly alter this figure".[36]

Skin — not routinely (Insufficient evidence)

The USPSTF gives an "I" statement to routine clinician-performed visual skin checks in asymptomatic adults.[37] A systematic review of 20 studies covering over 6 million people found no population-level melanoma mortality benefit — though the evidence bearing directly on mortality is thinner than that headline suggests: three non-randomised analyses of two German screening programmes, in about 1.8 million people, followed for four to ten years.[38] Germany has offered free biennial whole-body skin checks to every adult 35 and over since 2008, and melanoma mortality there fell no faster than in nine neighbouring countries with no screening at all — −1.8% a year against −2.2%.[39] Routine screening finds a great deal of Stage 0 / Stage I melanoma — much of which appears to be indolent — without proportionally reducing late-stage, lethal nodular melanoma, which often progresses rapidly between screening intervals (length-time bias).

This isn't an argument against dermatologic care for visible suspicious lesions or for patients with high-risk family history, very fair skin that burns rather than tans, or many atypical moles. It is an argument against the routine annual "skin check" as a population-wide longevity lever.

Hepatitis C and aortic aneurysm — one-time, defined-population screens (Strong)

Two screening tests deliver outsized benefit relative to their effort because they are one-time, well-targeted, and curative or definitively actionable downstream.

  • Hepatitis C, one-time for most adults 18–79. With direct-acting antivirals clearing the virus in 95% to 99% of people who take them, depending on genotype, the case for universal one-time hepatitis C virus (HCV) antibody screening — with an automatic follow-up test for viral genetic material if the antibody is positive — is now overwhelming.[40] The USPSTF dropped the prior birth-cohort and risk-based framing in 2020, and screens everyone 18 to 79 once, with periodic repeat testing for anyone at continued risk.[41] Hepatitis B remains risk-stratified.
  • Abdominal aortic aneurysm (AAA) — one-time ultrasound, men 65–75 who have ever smoked. The USPSTF criterion is narrow because that is where the survival benefit concentrates. The 2024 European Society for Vascular Surgery guidelines emphasize the same point: defer CT angiography and surgical planning until the operative threshold (about 55 mm in men, 50 mm in women) is met, and route patients to volume-adequate centers (at least 15 open and 15 endovascular repairs annually).[42]

The overdiagnosis problem — visible most in thyroid (Caution)

Overdiagnosis isn't a screening failure mode; it's a screening byproduct. The clearest illustration is thyroid cancer: global incidence has surged for three decades while mortality has stayed flat. Modeling using a validated papillary thyroid carcinoma microsimulation framework estimates that 72% to 94% of all US papillary thyroid cancer diagnoses between 1991 and 2019 were overdiagnoses — clinically silent lesions that would never have surfaced symptomatically.[43] That is roughly 550,000 to 730,000 overdiagnosed Americans over the period, many of whom went on to total thyroidectomy, radioactive iodine, and lifelong thyroid hormone replacement for a "cancer" that didn't behave like one. See Thyroid management for the active-surveillance reframe — which the 2025 American Thyroid Association guidelines bring into the guideline itself for selected low-risk thyroid cancer, though as a conditional recommendation resting on low-certainty evidence rather than a new standard of care.[44]

SiteEstimated overdiagnosisWhat it produces
Papillary thyroid72–94% of detected cases (US 1991–2019)Thyroidectomy; lifelong thyroid hormone replacement; risk of parathyroid injury and lasting calcium problems
DCIS (breast)~20% of all DCIS in a screened population (higher if only screen-detected DCIS is counted)Lumpectomy ± radiation ± endocrine therapy
Localized prostate (PSA-detected)23–42% of screen-detected cases (US modelling of 1985–2000 practice, before MRI-first biopsy and routine active surveillance)[45]Prostatectomy / radiation; incontinence; erectile dysfunction
Stage 0/I melanomaIndolent share of rising incidenceExcisional surgery; psychological burden

The defensive principle is to screen only where evidence supports the cadence, take overdiagnosis seriously when deciding what to do with an early-stage result, and prefer active surveillance over reflex resection when the evidence supports it (DCIS, low-grade prostate, sub-centimeter thyroid cancers).

Multi-cancer early detection (MCED) — promising, not ready (Weak / preliminary)

The biggest unsolved problem in cancer screening is that about two-thirds of cancer deaths come from cancers with no recommended screening test at all — 68.6% on the one published estimate, produced by authors five of whom work or worked for a company selling one of these tests.[46] Multi-cancer early detection (MCED) assays — blood tests that read chemical tags on tumor DNA shed into the bloodstream, of which Galleri is the most prominent — promise to find dozens of cancers from a single draw. A UK National Institute for Health and Care Research systematic review of how well these tests perform found 36 studies and no completed randomised trial; most were at high risk of bias, chiefly because people who tested negative were rarely followed up to see whether they had cancer after all. Galleri's sensitivity ranged from 21% to 66% across studies, at a specificity of 98.4–99.5%. The review synthesised these narratively rather than pooling them, and its searches closed in September 2023, so it predates everything else in this section.[47]

The first large interventional readout is the NHS-Galleri trial in over 142,000 asymptomatic 50–77-year-olds. Its primary results were presented at the 2026 American Society of Clinical Oncology meeting; the peer-reviewed publication is still pending, so the figures below come from a conference abstract and from the manufacturer's own account of its own trial.[48] The headlines are mixed:

  • Primary endpoint missed: no reduction in combined Stage III/IV cancer incidence across 12 pre-specified cancers — 706 cases in the intervention arm against 688 in the control arm (incidence rate ratio 1.03, 95% CI 0.92–1.14). Stage IV alone, a secondary endpoint, was about 14% lower, with an interval that only just excludes no effect (0.86, 0.744–0.998).
  • Stage shift in the right direction: 16% more Stage I/II diagnoses (647 against 559), and about four times as many screen-detected cancers (1,173 against 290) when the test was added to standard screening.
  • Fewer emergency-presentation diagnoses — 21% fewer in the abstract (225 against 286; GRAIL's own account says 25%), meaningful for both prognosis and acute healthcare cost.
  • Few false alarms: after three screening rounds only about 1.2% of the intervention arm were sent for investigation without a cancer being found, against a positive predictive value of 52%.[49]

The cautious read is that MCED tests find earlier-stage cancers without yet showing they prevent late-stage cancers — a pattern entirely compatible with detecting indolent disease faster rather than aggressive disease earlier. That distinction is the whole argument, and national screening bodies have made it explicit: a stage shift is a surrogate outcome, and a surrogate is only trustworthy once it has been shown to track the outcome that matters.[50] On that reasoning the UK National Screening Committee's 2026 position is that MCED tests need adequately powered randomized evidence of reduced cancer mortality before they can be recommended for population screening.[51] A fall in late-stage incidence is the most promising stand-in it will entertain, but its companion statement is explicit that surrogates "cannot yet replace mortality as the definitive trial outcome for benefit".[52] Professional oncology societies took the same line at ASCO in 2026; the USPSTF has issued no position on MCED at all. Insufficient evidence to recommend MCED for routine population screening outside clinical trials remains the consensus. That position will move when a trial reports a hard mortality endpoint, and there is a date on it: the NHS-Galleri protocol schedules its cancer-mortality analysis about two years after the stage readout, so around 2028.[53]

There is also a cost that test accuracy cannot capture. In the English regions that hosted the trial, the share of urgent referrals for suspected head and neck, lung and upper gastrointestinal cancer taking more than 28 days to reach a diagnosis rose while it fell everywhere else — by about 3 percentage points in the first six months and 5 in the next, an effect no longer significant after the first year.[54] Screening a healthy population competes for the same scanners and clinics that people with symptoms are waiting for. That is an argument about capacity rather than about the test, but it is the argument a health system actually has to settle.

Practical guidance

  1. Run the evidence-backed cadence and complete it. Colorectal from 45 — colonoscopy every 10 years or a stool test every year are the two most used, but stool DNA every 1 to 3 years and CT colonography every 5 years are equally covered, and the one you will actually do beats the one you keep deferring. Mammography biennially from 40 to 74. Primary HPV every 5 years from 30 to 65, or every 3 years if you collected the sample yourself. Low-dose CT annually if you meet the 50–80 / 20+ pack-year / ≤15 years quit criteria. One-time HCV antibody if you've never been tested. One-time AAA ultrasound for men 65–75 with any smoking history.
  2. Have the PSA conversation; don't reflex-order it. Shared decision around age 50 (45 with family history or African ancestry); use baseline-stratified intervals and ask about multiparametric MRI before a first biopsy if you do screen — the urology guideline makes it a conditional recommendation there and a strong one only before a repeat biopsy.[55]
  3. Take overdiagnosis seriously when you get a result. A Stage 0 finding (DCIS, ≤1 cm thyroid cancer, Gleason 6 prostate) doesn't automatically mean immediate surgery — active surveillance is an established option for low-grade prostate cancer, a guideline-endorsed but conditional one for selected low-risk thyroid cancer, and for low-risk DCIS a strategy that has now passed a two-year randomized non-inferiority test rather than a settled standard.
  4. Don't add tests that lack evidence. Routine whole-body skin checks, opportunistic neck ultrasound, full-body MRI marketed to healthy adults, and MCED panels outside trials all currently sit closer to harm than benefit on the population evidence.
  5. Recognize that screening reduces cancer mortality, not necessarily all-cause mortality. The mean life-extension in the trial data is small. The behavioral levers — not smoking, blood pressure, lipids, sleep, exercise, weight — still dominate the longevity arithmetic.

What's overrated

  • Whole-body MRI and CT for healthy adults. Marketed aggressively; no RCT evidence of mortality benefit; substantial false-positive cascade.
  • Routine annual dermatologic skin checks in average-risk asymptomatic adults — the population data show stage shift without lethal-melanoma mortality reduction.
  • Opportunistic neck ultrasound for nonpalpable thyroid nodules — the modality that produced an iatrogenic cancer epidemic.
  • MCED tests as currently sold — biologically impressive, not yet shown to reduce the cancers people actually die of.

Further reading

  • Bretthauer M et al. Estimated Lifetime Gained With Cancer Screening Tests: A Meta-Analysis of Randomized Clinical Trials. JAMA Intern Med 2023.[56]
  • Bretthauer M et al. Effect of Colonoscopy Screening on Risks of Colorectal Cancer and Related Death (NordICC). N Engl J Med 2022.[57]
  • Meirson T et al. Survival outcomes of population-wide colonoscopy screening: reanalysis of the NordICC data. 2024 — a restricted-mean-survival-time reanalysis of colorectal cancer incidence and colorectal cancer death, not of all-cause mortality.[58]
  • US Preventive Services Task Force. Screening for Breast Cancer: USPSTF Recommendation Statement. JAMA 2024.[59]
  • Ryser MD et al. Cancer outcomes in women without upfront surgery for ductal carcinoma in situ: observational cohort study. BMJ 2025.[60]
  • Poelhekken K et al. Overdiagnosis of ductal carcinoma in situ by grade and definition in population-based screening: a modeling study. The Breast 2025.[61]
  • Rebolj M et al. Primary cervical screening with high risk human papillomavirus testing: observational study. BMJ 2019.[62]
  • Roobol MJ et al. European Study of Prostate Cancer Screening — 23-Year Follow-up. N Engl J Med 2025.[63]
  • Tsodikov A et al. Reconciling the Effects of Screening on Prostate Cancer Mortality in the ERSPC and PLCO Trials. Ann Intern Med 2017.[64]
  • Henrikson NB et al. Skin Cancer Screening: Updated Evidence Report and Systematic Review for the USPSTF. JAMA 2023.[65]
  • Francis DO et al. Overdiagnosis of Papillary Thyroid Cancer. JAMA Netw Open 2026.[66]
  • Batson S et al. UK National Screening Committee position statement on evidence required for multicancer detection tests. BMJ 2026.[67]
  • NHS-Galleri trial — topline results. GRAIL press release, 2026.[68]

— § —